Method and device for monitoring a state of a component of a process engineering plant

Abstract

The invention relates to a method and a corresponding device for monitoring the state of a component of a process plant comprising a plurality of process-related interconnected components. The state of a component is determined based on process variables from at least one sensor relevant to the component, and corresponding data sets are recorded and stored for each sensor. According to the invention, a graph is generated from a digital flow diagram of the plant, which contains the structure of the plant with its components and their functions as additional information and their functional relationships. In this graph, the components of the plant are represented as nodes, and the functional relationships between the components are represented as lines of action according to a flow direction defined in the flow diagram, based on the additional information and relationships specified in the flow diagram.In the graph, a node is selected as the target node. The target node corresponds to a component whose state is to be monitored using a machine learning (ML) algorithm. Based on a rule and / or metric, input nodes of the previously selected target node are then identified, and sensor data sets from the input nodes that are relevant for monitoring the component being monitored, as well as the data sets from the relevant sensors of the input nodes, are selected to train the ML algorithm of the target node. Using this trained ML algorithm, the state determination of the component represented by the previously selected target node is significantly improved.

Description

[0001] The invention relates to a method for monitoring the condition of a component of a process plant according to the preamble of claim 1 and a corresponding device. Furthermore, the invention relates to a corresponding computer program and computer program product.

[0002] Condition monitoring refers to the regular or continuous recording of the technical condition of a machine or machine component using sensors, and the resulting sensor data is analyzed for further use. Meaningful sensor readings are usually physical variables such as temperature, pressure, vibration, humidity, fill levels, etc.

[0003] Condition monitoring is used particularly for process engineering components in complex process engineering plants in the process industry. In this context, the term (process engineering) component encompasses various machines or units within the plant, such as vessels, pumps, valves, heat exchangers, and their interconnections, particularly pipelines. All components serve to convey, separate, temper, react, or otherwise modify the composition, type, or properties of certain media such as liquids, gases, or solids. The sensors within a process engineering plant are designed to record process variables such as temperature, pressure, fill levels, etc.

[0004] The simplest form of condition monitoring is to examine every sensor reading associated with a device and analyze it for anomalies. This means checking whether the current value is within specified limits, which means the machine is OK. If the current value is outside the limits, the device is faulty, and an alarm is sent.

[0005] Machine learning-based methods are often used to detect anomalies in sensor data. In so-called "machine learning" (ML), an artificial, computer-implemented system learns patterns and relationships from data. In this way, after the learning phase, relationships can be generalized from examples based on historical data, and outliers and unknown patterns can be identified. To do this, algorithms build a statistical model based on training or learning data. If a machine learning algorithm (ML algorithm) is to be trained, for example, to monitor or predict the behavior of a process component in an industrial plant, the algorithm must first be fed a training data set.

[0006] In large and complex process plants, sensor data from a variety of sources often overlap. In many cases, the training dataset for a machine learning model contains data from sensors that are irrelevant to the behavior of the component being monitored. Furthermore, the values ​​recorded by the sensors are usually noisy. This complicates the training of the algorithm or model, so that patterns and relationships in the sensor data are often poorly recognized, and component monitoring leads to incorrect conclusions.

[0007] Improvements in condition monitoring using machine learning can be achieved by using extremely large datasets, particularly by improving training. However, such large amounts of data are not always available, and training with such a dataset would be correspondingly time-consuming and complicated. The quality of training (and evaluation) can be improved by manually selecting datasets from sensors that are not required to monitor a component in a technical system. The disadvantage of this is that it is labor-intensive. In addition, domain-specific knowledge of the technical system is often required to determine the relevant sensor datasets.In such cases, a person specializing in data analysis without sufficient overview of the process engineering in a technical plant is dependent on the collaboration of a person with domain knowledge, which, in addition to the large amount of work involved, represents a considerable additional burden for several people and is often not feasible for many projects.

[0008] The invention is based on the object of improving the condition monitoring of components of process engineering plants using ML algorithms and, in particular, of reducing the high workload in selecting training data sets for the ML algorithms used.

[0009] This object is achieved by the features of independent patent claim 1. The dependent claims describe advantageous developments of the invention, claim 8 describes a computer program, and claim 9 describes a computer program product.

[0010] The invention is based on the finding that the ML algorithms used for condition monitoring of components in technical systems deliver significantly better results when the structure of the system, as defined by a flow chart, is taken into account when training the ML models. The only prerequisite for this is the existence of a digital flow chart.

[0011] According to the invention, a graph is generated from the existing digital flow diagram of a process engineering plant, which contains the structure of the technical plant with its components and their functions as additional information and their functional interrelationships. In this graph, according to the additional information and interrelationships specified from the flow diagram, the components of the technical plant are contained as nodes and the functional interrelationships between the components are contained as lines of action according to a flow direction of the flow diagram. A node is selected in the graph as the target node. The target node corresponds to a component whose status is to be monitored using an ML algorithm. According to a rule and / or a metric related to the graph, input nodes of the previously selected target node are then determined, as are sensor data records of the input nodes that are relevant for monitoring the component to be monitored.The data sets from the relevant sensors of the input nodes are used as input data for training the ML algorithm of the target node. Using the output data, the results of the ML algorithm, the state of the component to be monitored is determined.

[0012] The main advantage of the invention is that the targeted specification of input data sets for the ML algorithm, which is based solely on the structure of the system, significantly improves the quality of the training of the ML algorithm and the use of the ML algorithm in the condition monitoring of components of technical systems. Domain-specific knowledge of the technical system is not required. By specifying a rule and / or metric, the laborious, manual selection of relevant sensor data sets is eliminated. After selecting the component to be monitored, which is represented as the target node in the graph representation, the process can run largely automatically and quickly and reliably delivers the output data of the ML algorithm. The result of the ML algorithm will usually be at least one process variable that characterizes the condition of a component of the technical system.

[0013] In a particularly advantageous embodiment, a user can define a rule and / or metric themselves, i.e., specify or select it. This allows for a high degree of application flexibility. For example, it is conceivable that a user selects specific metrics for the relationships between the target and input nodes using a menu in a graphical user interface. Or that the process variables of the input nodes recorded by specific sensors can be entered using an input menu. This embodiment allows the user to optimize the process, as different constellations of metrics or different rules can be entered to evaluate the calculation result of the ML algorithm.

[0014] In further advantageous embodiments of the method according to the invention, the data sets per sensor correspond either to measured values ​​of the process variables or to simulated values ​​of the process variables. Data sets of measured values ​​of the process variables recorded in real time allow condition monitoring of components during ongoing operation of the technical system. If simulated sensor data sets are used, the method according to the invention can be applied, for example, before commissioning of a technical system. In particular, a digital twin of the technical system could provide the data sets for monitoring the system, so that the ML algorithm can be trained using these simulated data sets outside of the technical system's operation.

[0015] In a further advantageous embodiment of the method according to the invention, the ML algorithm is trained in advance or during ongoing operation of the technical system. Extracting the training step allows for improved condition monitoring, since the actual condition determination is only performed with a trained algorithm, which leads to better calculation results.

[0016] It's particularly advantageous if the training time range is adjustable. This allows users to check for themselves when the ML algorithm delivers the best results.

[0017] The described developments relate both to the method according to the invention and to the device.

[0018] Furthermore, a realization of the invention or a described further development is possible by a computer program, in particular a software application, with computer-executable program code instructions for implementing the method when the computer program is executed on a computer.

[0019] The invention and / or any further development described can also be implemented by a computer program product which has a storage medium on which a computer program is stored which carries out the invention and / or the further development.

[0020] The computer program product can advantageously be transferred to the main memory of a computing unit and executed from there with the aid of at least one CPU. The computer program product can advantageously be stored on a data storage device such as a USB stick, a hard disk, or a CD-ROM / DVD-ROM and from there can be retrieved or installed on the computing unit.

[0021] In the following, the invention is described and explained in more detail with reference to the embodiments shown in the figures.

[0022] In it, each in a simplified schematic representation: FIG. 1 shows an exemplary section of a flow diagram of a process engineering plant. FIG. 2 shows a graph in accordance with a first exemplary embodiment of the present invention. FIG. 3 shows a graph in accordance with a second exemplary embodiment of the present invention. FIG. 4 shows two examples of the time courses of an estimated value of a process variable of a monitored component of a plant resulting from the ML algorithm. FIG. 5 shows an embodiment of a device for carrying out the method according to the invention for monitoring a state of a component of a process engineering plant.

[0023] A prerequisite for the present invention is the availability of a flow diagram in digital form. A flow diagram is a tool in the form of a technical drawing within process engineering. A flow diagram of a process engineering plant contains the structure of the plant with its components and their functions and functional relationships as additional information. One embodiment of such a flow diagram is the piping and instrumentation diagram (P&ID). It shows the connection of the lines, the flow directions, and all installations of a pipeline. Digitalized flow diagrams can be created using plant planning tools such as COMOS from Siemens AG.

[0024] Insofar as this application refers to technical systems and flow diagrams, this also includes individual, independent or interconnected (or to be interconnected) sub-systems or sub-flow diagrams.

[0025] FIG. 1 shows an example section of a piping and instrumentation flow diagram of a process engineering plant. The section FSA contains graphical process objects that represent the various components of the plant, such as pumps, vessels, butterfly valves or valves, and are linked to one another according to the piping of the plant. The process objects usually also contain technical information about the parameters of the components they represent, such as the pipe diameter or the filling volume of a vessel (not shown here). The flow diagram section FSA shown here shows a vessel B in which the fill level is measured using a sensor S1. The inlet and outlet of vessel B can be seen by lines of action with arrows. The outlet of the vessel branches into two branches.In the corresponding pipe sections there are pumps P1 and P2 with associated valves V1 and V2 and sensors S3 and S2 for measuring the flow rate at the pump outlet. Butterfly valves A1 and A2 are arranged downstream of pumps P1 and P2 in the direction of flow. Further in the flow direction of the common discharge pipe there is a pressure sensor S4 and another valve V3. In addition to the sensors, which are represented here by sensors S1, S2, S3 and S4, the actuators (represented by motors M for driving parts of the components) of the system with the corresponding technical information (not shown here) can be found in the flow diagram. Process variables such as pressure, temperature, fill level or flow rate are measured using the sensors. For each sensor there are corresponding data records, which can consist of time series data of the measured process variables, for example, and are stored in a buffer or a database.

[0026] From the FSA excerpt, it is clear that the state of a component (e.g., tank B, pumps P1 and P2, or butterfly valves A1 and A2) can be determined using a number of sensors. The component of pump P1 is directly causally related to sensors S1, S3, and S4. The causal relationships can be limited to one direction or bidirectional, i.e., acting in two directions. These causal relationships can now be represented in a simplified graphical representation.

[0027] An embodiment of a graph representation is shown in Figures 2 and 3. A graph within the meaning of the invention comprises any schematic representation in which, according to the additional information and relationships specified from the flow chart, the components of the technical system are contained as nodes K and the functional relationships between the components as lines of action W with a direction specified according to the flow direction of the flow chart or at least partially bidirectional. Lines of action according to a flow direction of the flow chart can therefore have one direction (for the flow of a medium through a pipe in one direction) or be bidirectional (in the case of a pressure equalization pipe). This can be a knowledge graph, for example. However, the invention is not limited to a knowledge graph.Different types of graphs or ontologies can be used as long as the type of objects and the connection between any two objects can be queried.

[0028] Graphs for representing causal relationships between objects or components of technical systems are widely known from the state of the art: US 2022 / 0075515 A1 discloses generating an interactive graphical user interface with multiple modes and panels that allow users to visualize and analyze models. A user can select an object from the graphical model and receive all linked objects with additional information. This information can be used to simulate the models.

[0029] WO 2022 / 035427 A1 describes a method for assisting an engineer in creating a CAD design of a technical process. An AI module is integrated into the engineering tool to classify the functionality of components for the current design project using a trained machine learning-based model. The AI ​​module receives a knowledge graph for the current project based on data associated with the graphical design of the technical process. The knowledge graph represents an ontology for a set of elements and element relationships that represent system components. The AI ​​module identifies the functionality for each node in the knowledge graph based on the classification model, groups the knowledge graph nodes according to the identified functionality, and generates functionality-based recommendations based on the groups in response to user queries.

[0030] WO 2020 / 198249 A1 describes systems and methods for controlling an industrial plant, comprising entering an engineering diagram for a unit of the industrial plant, the engineering diagram containing symbols representing assets of the industrial plant; extracting one or more assets from the engineering diagram using machine learning to recognize one or more assets, the one or more assets comprising equipment, instruments, connections, and lines, the lines connecting the equipment, instruments, and connections; determining one or more relationships between the equipment, instruments, connections, and lines using machine learning to recognize one or more relationships; and creating a flowchart from the assets and their relationships to each other.

[0031] In one example, the plant planning tool supports the export of the digital P&ID in the so-called DEXPI format. The abbreviation DEXPI stands for "Data Exchange in the Process Industry." The DEXPI format corresponds to a neutral data format that can be used to exchange neutral process information between software products from different engineering tool manufacturers. A digital flow chart in DEXPI format can be used as a graph itself or converted into another format. It contains all devices and sensors, as well as the connections between them and their directions, in a simplified graphical representation. The basis for the graphs is a common data model. When a P&ID is converted into a graph, a data transfer takes place that includes graphics, symbols, topology, all technical attributes, enumerations, selection lists, etc.to enable a seamless continuation of the P&ID functions in the target system.

[0032] If a digital P&I diagram is available, e.g. in the form of a knowledge graph or a similar ontology, it can then be determined which sensors in the respective application have a relevant reference to a device or component of a process plant to be monitored and are therefore particularly relevant for training an ML algorithm.

[0033] If the status of a specific device or component of a process plant is to be monitored using an ML algorithm, a user of the method according to the invention, which can be implemented as a computer program product, selects a node of the graph representation as the target node in one step. The target node within the graph representation corresponds to a component whose status is to be monitored. The selection of the target node can alternatively be automatic and / or specified by software. An ML algorithm is to be trained for the target node.

[0034] In the next step, the sensor data sets are to be determined which have a causal relationship with the component of the target node and thus determine the state of the component of the target node to be monitored. The components and sensors that are relevant for monitoring the component of the target node to be monitored are referred to as input nodes in the graph representation. These can be nodes upstream of the target node in the direction of flow, as well as nodes downstream of the target node in the direction of flow. Which input nodes (components) are relevant for determining the state of the component of the target node depends on the respective application. According to the invention, it is also possible to specify certain input nodes according to a metric or a rule.

[0035] For the purposes of this invention, a metric is understood to be a distance function or any distance measure, as defined mathematically, between two points in a space. This can be, for example, the length of the shortest path, resistance distance, or Euclidean distances. Essentially, a metric is intended to determine how "close" the (active) connection between the sensor and the device to be monitored must be for the sensor to be considered relevant. In the simplest case, the nearest neighboring nodes with their sensor data sets could be selected as input nodes. However, more complex cases will often arise from the application. For example, if the temperature is to be determined that will prevail at a component of the target node in 5 minutes, and the fluid transported in the supply pipes is approximately 1 +- 0.If the system takes 3 minutes to reach the next component, then it would make sense to select only the upstream nodes located between 4 and 6 nodes from the target node as input nodes. In this case, the input nodes would be selected according to a rule. Another example of a rule would be to select all pressure sensors located at a distance of no more than one node downstream as input nodes, as well as all upstream temperature sensors of the nodes located 2 to 3 nodes from the target node.

[0036] In a particularly advantageous embodiment of the invention, the rules and metrics that determine which nodes of the graph representation are to be defined as relevant input nodes for the corresponding application case can be specified by a user (a data analyst or a domain expert). The rules and metrics that determine which nodes are considered relevant for the specific problem can, for example, be established once and can subsequently be applied to all sufficiently similar devices / target nodes (e.g., to all similar pumps in the system). This allows all sufficiently similar devices / target nodes, and thus the entire system, to be monitored in this regard during ongoing operation without significant additional effort.

[0037] Since each node of the graph represents a component of the process plant that is connected to various sensors (see also Fig. 1 ), the sensor data sets relevant for determining the state of the component of the target node result from the determination of the input nodes, which are used as input data for training the ML algorithm of the target node, whereby the state of the target node is determined using the output data of the ML algorithm.

[0038] In Fig. 2 A first embodiment is given to illustrate the method according to the invention. In the graph representation in Fig. 2 , which is based on a P&ID flow diagram of a process plant, certain components required for the operation of the process plant are represented as nodes K and the pipes running between the components are represented as lines of action. In the highly simplified section of the P&ID flow diagram shown as a graph, the flow direction runs along the arrow directions, generally from top to bottom with diagonal connections to the nearest neighbors, although some connections are missing. In this case, the components of the nodes represent temperature measuring points within a piping system. The nodes therefore correspond to a pipe section at which the temperature of the transported liquid is measured with a sensor (temperature sensor). The arrows between the nodes indicate the flow direction of the liquid between pipe sections with the sensors.In this embodiment, the component and sensor are therefore considered to be the same. In general, each component of a node in the graph representation is connected to more than one sensor.

[0039] In this exemplary embodiment, an ML algorithm is to be trained for the target node K43, which is to monitor the condition of the pipeline at this measuring point by estimating the temperature in this pipe section using the ML algorithm. The measuring point within the piping diagram is represented by the target node K43. Due to the circumstances of this application, only the data sets from the upstream temperature sensors are relevant for determining the condition of the pipe section at the target node, since only these contain information about the fluid flowing into the pipe section represented by the target node. Furthermore, temperature sensors located far upstream can be ignored, since their information is reflected in the data sets of the later sensors and is heavily mixed with the information from irrelevant sensors before the target sensor is reached.In this example, a maximum distance of two nodes is chosen between the target node and the input nodes determined to be relevant. The metric here is therefore the length of the shortest path. In . Fig. 2 The input nodes K32, K22, K33, K23, and K24 of the target node K43 can be seen in the dark gray hatched area. Only the data sets from the sensors of the input nodes K32, K22, K33, K23, and K24 are used to train the ML algorithm of the target node. The data sets from the more distant sensors of nodes K00 to K16 in the light gray area do not need to be used to train the ML algorithm and are negligible for the condition of the pipe section represented by the target node. The data sets related to the nodes located downstream of the target node are irrelevant for determining the condition of the target node.

[0040] In Fig. 3 A second exemplary embodiment is provided to illustrate the method according to the invention. In this graph, the nodes represent pipe sections within a pipeline system, at which both the temperature of the transported liquid is measured with one sensor (temperature sensor), and the pressure within the pipe section is measured with another sensor. The arrows between the nodes also indicate the direction of flow of the liquid between the pipe sections with the sensors. In this exemplary embodiment, each component (=pipe section) is connected to two sensors (here, a temperature and a pressure sensor).

[0041] In this exemplary embodiment, node K33 is selected as the target node for which an ML algorithm is to be trained, which is to monitor the condition of the pipeline in this pipe section. The input nodes are selected according to a rule. For the ML algorithm, only the data sets from the pressure sensors of the neighboring nodes (i.e., the nodes at a distance of at most one node from the target node) and the data sets from the temperature sensors of the upstream nodes at a distance of 2 to 3 nodes are to be considered. Thus, nodes K22, K23, K24, K43, and K44 are determined as input nodes, whereby only the data sets from the pressure sensors of these nodes are relevant for determining the condition of the pipe section of the target node K33. In the graph representation in Fig. 3 These are the nodes in the light grey field P. In addition, the nodes in the dark grey field T are also determined as input nodes, whereby only the data records of the temperature sensors are taken into account.

[0042] The selection of only certain, relevant data sets as input data sets for an ML algorithm according to the method according to the invention contributes to a significant improvement of the output value calculated by the algorithm. This is shown in the two diagrams in Fig. 4 Both diagrams show the temporal progression of the output variable of the ML algorithm for the process value of the component of the target node K43 from Example 1 (cf. Fig. 2 ). The node K43 from Fig. 2 represents a temperature sensor in a pipe section of a piping system. Using the ML algorithm, the measured temperature value can be estimated based on the input data sets, i.e., the data sets of the input nodes. In both diagrams D1 and D2, the output variable A of the ML algorithm used is plotted against time. Additionally, the target or setpoint value of the output variable S is plotted against time. In both diagrams, the time period TR spent training the algorithm is also marked.

[0043] Diagram D1 shows that during the training phase TR, the two curves of the output variable, i.e., the target values ​​and the temperature values ​​calculated using the ML algorithm, are still approximately the same. In the evaluation phase, however, significant deviations occur in D1. This is due to the fact that the training data sets for the ML algorithm in diagram D1 also contain irrelevant input data sets, while in diagram D2 only relevant training data sets were used to calculate the estimated temperature values ​​according to the invention. In diagram D1, the ML algorithm was tested using the data sets of all nodes (60 temperature sensors) of the Fig. 2 shown graph representation. In diagram D2, the ML algorithm was trained only with the relevant data sets of nodes K22, K23, K24, K32 and K33 of the Fig. 2 The graph shown is trained. In this way, a significant improvement in the condition monitoring of the pipe section in question can be achieved.

[0044] The invention and the described developments are preferably realized in software as well as in hardware, for example using a special electrical circuit or implemented in a combination of software and hardware such as a data processing device.

[0045] In Fig. 5an embodiment of a device CM for monitoring the status of a component of a process engineering plant is shown. In this embodiment, the device CM has at least one interface S1 for receiving a digital flow diagram FS of the technical plant. Such a flow or piping diagram (P&ID diagram), which may consist of several partial images, is usually created using a plant planning tool APW in a known manner, e.g. by a plant planner or technologist, by linking graphic process objects representing the components of the technical plant. The device CM further has at least one further interface S2 for receiving data records D1 to DN (N = natural number) or variables derived therefrom from sensors S1 to SN, which are connected to at least one component of the process engineering plant.The data sets containing process variables acquired by the sensors can be temporarily stored in a database DB. Alternatively, the device CM can also have a single interface configured to receive any data and / or files.

[0046] The device CM further comprises at least one data processing device DV, which executes the method according to the invention based on the transmitted digital flow chart and the transmitted sensor data sets. The device CM can further comprise an archive or any other storage unit SP or database in which, for example, the flow charts or graphs are stored. The data processing device DV has at least one processor P or CPU and a working memory M. The invention implemented as a computer program can, for example, be stored in the working memory M or loaded into it and executed from there with the aid of the at least one processor P.

[0047] The device may further comprise or be connected to a display unit A configured to monitor a component of the technical system on a user interface (GUI). A user can interact with the data processing device DV of the device CM as desired via the graphical user interface.

Claims

1. Method for monitoring a condition of a component of a process engineering plant, which has a large number of components that are connected to one another from a process engineering perspective, - wherein the condition of a component is determined on the basis of process variables of at least one sensor relevant for the component and corresponding data sets are recorded and stored for each sensor, - wherein from a digital flow diagram of the technical plant, which contains the structure of the technical plant with its components and the functions thereof as additional information and the functional relationships thereof, a graphic is generated which, according to the additional information and relationships specified from the flow diagram, contains the components of the technical plant as nodes and the functional relationships between the components as working lines, according to a flow direction of the flow diagram, characterised in that - a node of the graphic is selected as target node, which corresponds to a component, the condition of which is to be monitored by means of an ML algorithm, - according to a rule and / or a metric related to the graphic, input nodes and sensor data sets of the input nodes, which are relevant for the monitoring of the component to be monitored, of the target node selected in advance are ascertained, - the data sets of the relevant sensors of the input nodes are selected as input data for the training for the ML algorithm of the target node, and - by means of the output data of the ML algorithm, the condition of the component to be monitored is determined.

2. Method according to claim 1, characterised in that the rule and / or metric is specified or selected by an operator.

3. Method according to one of the preceding claims, characterised in that the data sets for each sensor correspond to measurement values of the process variables or simulated values of the process variables.

4. Method according to one of the preceding claims, characterised in that the training of the ML algorithm takes place in advance or during ongoing operation of the technical plant.

5. Method according to one of the preceding claims, characterised in that it is possible to set a time range of the training.

6. Apparatus (CM) for monitoring a condition of a component of a process engineering plant, which has a large number of components that are connected to one another from a process engineering perspective, wherein each component is connected to a number of sensors (S1, ...SN) for the detection of data sets (D1, ...DN) of process variables, - with an interface (S1) for receiving a digital flow diagram (FS) of the technical plant, - with a further interface (S2) for receiving the data sets (D1, ...DN) or variables derived therefrom of the sensors (S1, ...SN) of the components of the technical plant, and - with a data processing facility (DV), which is embodied to perform the method according to one of claims 1 to 5.

7. Apparatus (CM) according to claim 6, further having a display unit (A), which is embodied such that by means of a user interface (GUI) at least one selection of a target node takes place, which is related to the component that is to be monitored, and such that any given interactions of an operator with the data processing facility (DV) can be performed.

8. Computer program, in particular software application, with program code instructions that can be carried out by a computer in order to implement the method according to one of claims 1 to 5, when the computer program is executed on a computer.

9. Computer program product, in particular data carrier or storage medium, with a computer program according to claim 8 that can be executed by a computer.

Images